https://peterponyu.github.io/scCCVGBen/methods/Recent content in Methods on scCCVGBen — single-cell scCCVGBen BenchmarkHugo -- gohugo.ioen-ushttps://peterponyu.github.io/scCCVGBen/methods/gat/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/gat/GAT # Field Value Family attention Group scCCVGBen graph encoder Description # Graph Attention Network (Veličković 2018) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_GAT. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/gatv2/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/gatv2/GATv2 # Field Value Family attention Group scCCVGBen graph encoder Description # Dynamic attention Graph Attention v2 (Brody 2022) — scCCVGBen extension Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_GATv2. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/transformer/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/transformer/Transformer # Field Value Family attention Group scCCVGBen graph encoder Description # TransformerConv — Attention is All You Need graph variant (Shi 2020) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_Transformer. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/supergat/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/supergat/SuperGAT # Field Value Family attention Group scCCVGBen graph encoder Description # Self-supervised edge prediction GAT (Kim 2020) — scCCVGBen extension Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_SuperGAT. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/gcn/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/gcn/GCN # Field Value Family message-pass Group scCCVGBen graph encoder Description # Graph Convolutional Network (Kipf 2017) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_GCN. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/sage/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/sage/SAGE # Field Value Family message-pass Group scCCVGBen graph encoder Description # GraphSAGE inductive (Hamilton 2017) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_SAGE. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/graph/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/graph/Graph # Field Value Family message-pass Group scCCVGBen graph encoder Description # GraphConv — general message-passing variant Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_Graph. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/cheb/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/cheb/Cheb # Field Value Family message-pass Group scCCVGBen graph encoder Description # ChebNet — Chebyshev polynomial filters (Defferrard 2016) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_Cheb. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/tag/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/tag/TAG # Field Value Family message-pass Group scCCVGBen graph encoder Description # Topology-Adaptive Graph Conv (Du 2017) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_TAG. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/arma/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/arma/ARMA # Field Value Family message-pass Group scCCVGBen graph encoder Description # ARMA filter Conv (Bianchi 2021) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_ARMA. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/sg/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/sg/SG # Field Value Family message-pass Group scCCVGBen graph encoder Description # Simplified Graph Conv (Wu 2019) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_SG. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/ssg/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/ssg/SSG # Field Value Family message-pass Group scCCVGBen graph encoder Description # Simple Spectral Graph Conv (Zhu 2021) Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_SSG. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/gin/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/gin/GIN # Field Value Family message-pass Group scCCVGBen graph encoder Description # Graph Isomorphism Network (Xu 2019) — scCCVGBen extension Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_GIN. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/edgeconv/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/edgeconv/EdgeConv # Field Value Family message-pass Group scCCVGBen graph encoder Description # Dynamic Edge Conv (Wang 2019) — scCCVGBen extension Role in scCCVGBen # Axis A (encoder-variation) sweep: scCCVGBen trains a latent representation with this message-passing / attention module while holding the graph fixed to k-NN Euclidean. Benchmark naming for sweep rows: scCCVGBen_EdgeConv. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/knn-euclidean/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/knn-euclidean/kNN_euclidean # Field Value Family graph Group Graph construction method Description # Standard k-NN with Euclidean distance (k=15) — scCCVGBen benchmark default Role in scCCVGBen # Axis B (graph-construction sweep): scCCVGBen encoder is fixed to GAT while this graph builder constructs the cell-cell neighbourhood fed to the encoder. Benchmark naming: scCCVGBen_GAT_kNN_euclidean. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/knn-cosine/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/knn-cosine/kNN_cosine # Field Value Family graph Group Graph construction method Description # k-NN with cosine similarity — rewards direction, invariant to magnitude Role in scCCVGBen # Axis B (graph-construction sweep): scCCVGBen encoder is fixed to GAT while this graph builder constructs the cell-cell neighbourhood fed to the encoder. Benchmark naming: scCCVGBen_GAT_kNN_cosine. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/snn/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/snn/snn # Field Value Family graph Group Graph construction method Description # Shared Nearest Neighbour — 2 cells connected if they share a fraction of neighbours Role in scCCVGBen # Axis B (graph-construction sweep): scCCVGBen encoder is fixed to GAT while this graph builder constructs the cell-cell neighbourhood fed to the encoder. Benchmark naming: scCCVGBen_GAT_snn. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/mutual-knn/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/mutual-knn/mutual_knn # Field Value Family graph Group Graph construction method Description # Mutual k-NN — only edges where both cells are in each other’s k-NN list; stricter connectivity Role in scCCVGBen # Axis B (graph-construction sweep): scCCVGBen encoder is fixed to GAT while this graph builder constructs the cell-cell neighbourhood fed to the encoder. Benchmark naming: scCCVGBen_GAT_mutual_knn. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/gaussian-threshold/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/gaussian-threshold/gaussian_threshold # Field Value Family graph Group Graph construction method Description # Gaussian heat-kernel weights w=exp(-d²/(2σ²)); edges pruned at threshold 0.9 Role in scCCVGBen # Axis B (graph-construction sweep): scCCVGBen encoder is fixed to GAT while this graph builder constructs the cell-cell neighbourhood fed to the encoder. Benchmark naming: scCCVGBen_GAT_gaussian_threshold. scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/pca/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/pca/PCA # Field Value Family baseline Group Dimensionality-reduction baseline Description # Linear PCA — sklearn.decomposition.PCA Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: PCA (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/kpca/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/kpca/KPCA # Field Value Family baseline Group Dimensionality-reduction baseline Description # Kernel PCA (RBF) — sklearn.decomposition.KernelPCA Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: KPCA (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/ica/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/ica/ICA # Field Value Family baseline Group Dimensionality-reduction baseline Description # Independent Component Analysis (FastICA) — sklearn.decomposition.FastICA Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: ICA (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/fa/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/fa/FA # Field Value Family baseline Group Dimensionality-reduction baseline Description # Factor Analysis — sklearn.decomposition.FactorAnalysis Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: FA (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/nmf/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/nmf/NMF # Field Value Family baseline Group Dimensionality-reduction baseline Description # Non-negative Matrix Factorisation — sklearn.decomposition.NMF Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: NMF (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/tsvd/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/tsvd/TSVD # Field Value Family baseline Group Dimensionality-reduction baseline Description # Truncated SVD — sklearn.decomposition.TruncatedSVD Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: TSVD (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/dicl/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/dicl/DICL # Field Value Family baseline Group Dimensionality-reduction baseline Description # Dictionary Learning — sklearn.decomposition.DictionaryLearning Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: DICL (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/scvi/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/scvi/scVI # Field Value Family baseline Group Dimensionality-reduction baseline Description # Single-cell Variational Inference — Lopez 2018 Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: scVI (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/dip/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/dip/DIP # Field Value Family baseline Group Dimensionality-reduction baseline Description # DIP-VAE disentangled autoencoder — Kumar 2017 Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: DIP (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/info/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/info/INFO # Field Value Family baseline Group Dimensionality-reduction baseline Description # InfoVAE — Zhao 2017 Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: INFO (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/tc/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/tc/TC # Field Value Family baseline Group Dimensionality-reduction baseline Description # β-TCVAE — Chen 2018 Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: TC (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/highbeta/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/highbeta/highBeta # Field Value Family baseline Group Dimensionality-reduction baseline Description # Hyper-parameterised VAE with high β (β=100) Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: highBeta (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.https://peterponyu.github.io/scCCVGBen/methods/scccvgben/Mon, 01 Jan 0001 00:00:00 +0000https://peterponyu.github.io/scCCVGBen/methods/scccvgben/scCCVGBen # Field Value Family baseline Group Dimensionality-reduction baseline Description # Core scCCVGBen reference row; encoder and graph-axis variants are labelled separately Role in scCCVGBen # Axis C (baseline comparison): this method produces a latent embedding evaluated with the same curated 20 publication-display metrics as scCCVGBen. Benchmark naming: scCCVGBen (row label is the method name itself, with no scCCVGBen prefix). scCCVGBen benchmark tests each method against the same 200 public dataset records and curated 20 publication-display metrics; see the Methods index for the full set.